Mid-air gestures have no contact click to close the loop

What it is

A key has travel and a bottom. A touchscreen deforms skin. A mouse has a switch. Mid-air motion removes all of that contact feedback: no surface tells the finger “this hit happened.” Proprioception still knows where the arm is. It does not know whether the system harvested the motion as an event. What is missing is not decorative vibration. It is the physical edge that closes the action. Substitutes have to be found on another channel: vision, sound, artificial haptics, or self-contact such as a pinch. Without a substitute, people infer from consequences (a light, a skipped track). Delay the consequence and retries and double fires arrive.

Why it happens

Cutaneous mechanoreceptors are extremely sensitive to collisions and force steps; a key click pins the firing instant at about 10 ms. Mid-air has no step. The firing instant has to be aligned with a visual change or a sound, and visual feedback still stacks display delay. Proprioception reports joint angles, not “the outside world took it.” The loop is missing its hardest link. A pinch restores a patch of self-contact at the pads, so it is easier than a wave to know “I pressed”; it still does not know “the device took it.” A device-side substitute must arrive at the same instant as the commit. Later than the retry window, it fails.

Studying it

Wire the same command to: no substitute (consequences only), a visual pulse, a sound, a wrist buzz, fingertip self-contact. Measure time from action end to “I have clicked,” retry count, and double fires. Delay as a factor: hold feedback back 0–300 ms. Control with a physical key. Do not use preference items alone. “I like the buzz better” does not mean the loop was closed; retries and double fires do.

Where it stops holding

When the consequence is itself fast and at the point of gaze (a button under the cursor depresses at once), visual consequence can impersonate contact feedback. That needs near-zero delay and high contrast, which a living-room television does not have. Sound dies in noise. A buzz in a headset rather than in the commanding hand is spatially misaligned and gets read as a system notification, not this hit. Self-contact covers pinch, not swipe or palm-flip. Totally blind users need a non-visual substitute; “watch the screen flash” is not a solution.

Applying it

• For every mid-air action that commits a system event, name a substitute edge that arrives at the commit instant. Ban consequence-only closure.
• Budget feedback delay under the time at which people start to retry. If you cannot hit it, do not make that action the main path.
• Accept substitutes on retry count and double fires, not on “users said there was feedback.”

Related

• Same group: C4.11.2 In-range, locked, and fired must be separately legible · C4.11.3 Continuous action needs continuous feedback, not only an end bell · C4.11.4 Primary states need at least one non-visual channel
• Adjacent: C4.06 Air pinch · C4.26 Distinguishing three-state feedback
• Search: contact feedback · haptic click · mid-air confirmation